DOI: 10.1136/bmjinnov-2026-001575 ISSN: 2055-8074

Voice-driven clinical decision support for lipid management: integrating trial eligibility assessment with AI-powered guideline synthesis

Rugved Parmar, Daoud Eldawud, Adam Budzikowski

Background

Despite the proven efficacy of lipid-lowering therapy, only 30%–60% of eligible high-risk patients receive appropriate statin therapy. Guideline complexity and limited point-of-care access to trial eligibility criteria contribute to treatment gaps.

Objective

To develop a clinical decision support tool integrating structured clinical trial data with artificial intelligence for personalised lipid management recommendations.

Methods

We developed a web-based tool using Python/Streamlit incorporating nine landmark trials. The tool employs voice-driven input via OpenAI Whisper, GPT-4 for natural language processing, quantitative trial matching algorithms and comparative analysis of three international guidelines.

Results

The tool successfully processes natural language patient descriptions, provides granular trial eligibility assessments with specific inclusion/exclusion criteria analysis and generates evidence-based treatment recommendations with associated Class of Recommendation and Level of Evidence designations. Testing across four simulated clinical scenarios—post-ACS (Acute coronary syndromes) with diabetes, primary prevention with diabetes, statin intolerance with ASCVD (Atherosclerotic cardiovascular disease) and residual hypertriglyceridaemia—demonstrated the tool’s ability to identify differential trial applicability and surface alternative evidence pathways.

Conclusions

This proof of concept demonstrates a novel approach to clinical decision support through quantitative trial matching and AI-powered guideline synthesis. Formal validation is required before clinical implementation.

More from our Archive